Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement pairwise sum #548

Open
wolfv opened this issue Nov 21, 2017 · 3 comments
Open

Implement pairwise sum #548

wolfv opened this issue Nov 21, 2017 · 3 comments

Comments

@wolfv
Copy link
Member

wolfv commented Nov 21, 2017

When NumPy's sum function is called, it is actually a pairwise sum implementation, which is supposed to be more stable: https://en.wikipedia.org/wiki/Pairwise_summation

@wolfv
Copy link
Member Author

wolfv commented Feb 27, 2018

Actually, according to this, the pairwise sum doesn't seem to perform much better in terms of accuracy

https://github.com/nschloe/accupy

which questions if we should implement it. I vote for no (except we want to be 100% true to numpy).

@iamthebot
Copy link
Contributor

Assuming multithreaded single-axis reductions are implemented (eg; sum) you already get many of the benefits of pairwise summation (each thread computes a partial sum and then the partial sums are combined).

@wolfv
Copy link
Member Author

wolfv commented Jan 5, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants